Faster Equilibration?

[ASSIDUO Network]

Hi there everyone, so I am trying to run some AIMD simulations. Originally I used the NVT ensemble alongside the Nose-Hoover thermostat, but I was getting only 50 fs of simulation time done every day. I thus decided to experiment with different combinations of ensembles and thermostats, changing parameters for each where necessary. I ultimately found that the Langevin ensemble and adaptive-Langevin thermostat worked the best for me in my test runs. It got 562 fs of simulation time done every day, which is 10 times the speed improvement from my previous runs. This made me very excited.

I then decided to continue the simulation with this ensemble-thermostat combination. From the first restart (over 24 hours), I only got 52 fs of simulation time and now for the second restart (over 24 hours), I got less than 30 fs of simulation time. I am not sure why such a massive decrease happened when the memory and CPU usage has stayed similar.

Is there any way you know of that works to get a fast equilibration? 

[Jürg Hutter]

Hi

if you investigate the timing outputs at the end of your jobs, is there an obvious source
of the slowdown?
Are all routines slower or does one routine stand out (time/number of call)?

regards

Juerg Hutter

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of ASSIDUO Network <lenardc...@gmail.com>
Sent: Wednesday, May 25, 2022 8:16 AM
To: cp2k
Subject: [CP2K:17025] Faster Equilibration?

Hi there everyone, so I am trying to run some AIMD simulations. Originally I used the NVT ensemble alongside the Nose-Hoover thermostat, but I was getting only 50 fs of simulation time done every day. I thus decided to experiment with different combinations of ensembles and thermostats, changing parameters for each where necessary. I ultimately found that the Langevin ensemble and adaptive-Langevin thermostat worked the best for me in my test runs. It got 562 fs of simulation time done every day, which is 10 times the speed improvement from my previous runs. This made me very excited.

I then decided to continue the simulation with this ensemble-thermostat combination. From the first restart (over 24 hours), I only got 52 fs of simulation time and now for the second restart (over 24 hours), I got less than 30 fs of simulation time. I am not sure why such a massive decrease happened when the memory and CPU usage has stayed similar.

Is there any way you know of that works to get a fast equilibration?

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/1af99329-2cf3-440e-86fd-36b5718b26fan%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/1af99329-2cf3-440e-86fd-36b5718b26fan%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Lenard Carroll]

Hi there and thank you for the response. Unfortunately I am a bit lost in what you are asking me to check. I have included my output and restart files, in case you want to have a look yourself.

[Jürg Hutter]

Hi

whenever possible, you shouldn't stop runs by killing the process.

CP2K allows for a soft stop by creating in the running folder an empty file called EXIT.

This way you get full restart possibilities and a complete output file with all timing information.

regards

Juerg

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Jürg Hutter <hut...@chem.uzh.ch>
Sent: Wednesday, May 25, 2022 2:06 PM
To: cp...@googlegroups.com
Subject: Re: [CP2K:17030] Faster Equilibration?

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZR0P278MB07596BF7ED9B42A75FD45C239FD69%40ZR0P278MB0759.CHEP278.PROD.OUTLOOK.COM.

[Lenard Carroll]

I don't kill it, the walltime was reached for all the runs. My goal was to see roughly how many steps are required every 24 hours, then make sure I get to those steps before the walltime is reached. Unfortunately, it went from 530 fs, to 60 fs to 30 fs. The fourth run was 200 fs. It's not really easy to predict how many steps will be required.

Is there a way to do the soft stop at walltime limit?

You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/q2_FGSA07PQ/unsubscribe.
To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZR0P278MB0759ABA07503A422A26CE7929FD99%40ZR0P278MB0759.CHEP278.PROD.OUTLOOK.COM.

[Lenard Carroll]

If I use WALLTIME under GLOBAL, does it ensure a proper stop that will still have the timings?

[Jürg Hutter]

Yes, that is the way to do it. Give it a small save time (x*time for a SCF iteration).


________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Lenard Carroll <lenardc...@gmail.com>
Sent: Thursday, May 26, 2022 1:31 PM
To: cp...@googlegroups.com
Subject: Re: [CP2K:17047] Faster Equilibration?

If I use WALLTIME under GLOBAL, does it ensure a proper stop that will still have the timings?

On Thu, May 26, 2022 at 1:28 PM Lenard Carroll <lenardc...@gmail.com<mailto:lenardc...@gmail.com>> wrote:
I don't kill it, the walltime was reached for all the runs. My goal was to see roughly how many steps are required every 24 hours, then make sure I get to those steps before the walltime is reached. Unfortunately, it went from 530 fs, to 60 fs to 30 fs. The fourth run was 200 fs. It's not really easy to predict how many steps will be required.

Is there a way to do the soft stop at walltime limit?

On Thu, May 26, 2022 at 12:38 PM Jürg Hutter <hut...@chem.uzh.ch<mailto:hut...@chem.uzh.ch>> wrote:
Hi

whenever possible, you shouldn't stop runs by killing the process.

CP2K allows for a soft stop by creating in the running folder an empty file called EXIT.

This way you get full restart possibilities and a complete output file with all timing information.

regards

Juerg

________________________________________

From: cp...@googlegroups.com<mailto:cp...@googlegroups.com> <cp...@googlegroups.com<mailto:cp...@googlegroups.com>> on behalf of Jürg Hutter <hut...@chem.uzh.ch<mailto:hut...@chem.uzh.ch>>

Sent: Wednesday, May 25, 2022 2:06 PM

To: cp...@googlegroups.com<mailto:cp...@googlegroups.com>

Subject: Re: [CP2K:17030] Faster Equilibration?

Hi

if you investigate the timing outputs at the end of your jobs, is there an obvious source
of the slowdown?
Are all routines slower or does one routine stand out (time/number of call)?

regards

Juerg Hutter

________________________________________

From: cp...@googlegroups.com<mailto:cp...@googlegroups.com> <cp...@googlegroups.com<mailto:cp...@googlegroups.com>> on behalf of ASSIDUO Network <lenardc...@gmail.com<mailto:lenardc...@gmail.com>>

Sent: Wednesday, May 25, 2022 8:16 AM
To: cp2k
Subject: [CP2K:17025] Faster Equilibration?

Hi there everyone, so I am trying to run some AIMD simulations. Originally I used the NVT ensemble alongside the Nose-Hoover thermostat, but I was getting only 50 fs of simulation time done every day. I thus decided to experiment with different combinations of ensembles and thermostats, changing parameters for each where necessary. I ultimately found that the Langevin ensemble and adaptive-Langevin thermostat worked the best for me in my test runs. It got 562 fs of simulation time done every day, which is 10 times the speed improvement from my previous runs. This made me very excited.

I then decided to continue the simulation with this ensemble-thermostat combination. From the first restart (over 24 hours), I only got 52 fs of simulation time and now for the second restart (over 24 hours), I got less than 30 fs of simulation time. I am not sure why such a massive decrease happened when the memory and CPU usage has stayed similar.

Is there any way you know of that works to get a fast equilibration?

--
You received this message because you are subscribed to the Google Groups "cp2k" group.

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k%2Bunsu...@googlegroups.com><mailto:cp2k+uns...@googlegroups.com<mailto:cp2k%2Bunsu...@googlegroups.com>>.

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/1af99329-2cf3-440e-86fd-36b5718b26fan%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/1af99329-2cf3-440e-86fd-36b5718b26fan%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
You received this message because you are subscribed to the Google Groups "cp2k" group.

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k%2Bunsu...@googlegroups.com>.

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZR0P278MB07596BF7ED9B42A75FD45C239FD69%40ZR0P278MB0759.CHEP278.PROD.OUTLOOK.COM.

--
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/q2_FGSA07PQ/unsubscribe.

To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k%2Bunsu...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZR0P278MB0759ABA07503A422A26CE7929FD99%40ZR0P278MB0759.CHEP278.PROD.OUTLOOK.COM.

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAKur15_%2BR_u1wnqu5ypQ9xWXG2xYsv7-Ssx57rsOjDneRV5SsQ%40mail.gmail.com<https://groups.google.com/d/msgid/cp2k/CAKur15_%2BR_u1wnqu5ypQ9xWXG2xYsv7-Ssx57rsOjDneRV5SsQ%40mail.gmail.com?utm_medium=email&utm_source=footer>.

[Lenard Carroll]

Thank you. I will do this and if necessary, post the difference in timings if the problem still persists.

To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZR0P278MB075985E05DE8F9A12FC8B3619FD99%40ZR0P278MB0759.CHEP278.PROD.OUTLOOK.COM.